# Zscore Model Tutorial > 來源：https://uqer.io/community/share/54ab4407f9f06c276f6519ec ## 1. 什么是 Zscore Model 信用風險評分方法通常包括定性法、單變量法、多變量法，目前廣泛采用多變量方法包括如判別分析(discriminant analysis)、邏輯回歸(logit regression)和非線性模型如神經網絡等。奧特曼博士于1968年發表的Z-score模型基于多變量判別分析方法。 Z-score模型基于各變量加權得分對企業是否破產進行判斷，在Z-score原始模型中，得分高于2.99的屬于“安全”區域、低于1.80的屬于“困境”區域，兩個得分之間的屬于“灰色”區域。 Z-score模型在美國企業如柯達、通用汽車的破產預測得到了很好的結果。 ## 2. 本模塊提供的 Zscore Model 下面我們從原理，模型公式，劃分區間三個方面來簡單介紹一下本模塊中的兩個Z-score模型： ### 2.1 All Corporate Bonds 模型原理 + No equity prices are needed + Uses discriminant analysis methodology + Coefficients are obtained from China distressed firm dataset from historical periods 模型公式 ``` ZScore = 0.517 - 0.460*TotalLiabilities/TotalAssets + 9.320*NetProfit/0.5*(TotalAssets+TotalAssets[last]) + 0.388*WorkingCapital/TotalAssets + 1.158*RetainedEarnings/TotalAssets = 0.517 - 0.460*x1 + 9.320*2/x2 + 0.388*x3 + 1.158*x4 ``` 劃分區間 + `Z-score < 0.5`: 已經違約 + `0.5 < Z-score < 0.9`: 有違約的可能性 + `Z-score > 0.9`: 財務健康，短期內不會出現違約情況 ### 2.2 Corporate Bonds with Equity Listings 模型原理 + Uses Equity Prices: Information from equity set + Uses discriminant analysis methodology + Coefficients are obtained from China distressed firm dataset from historical periods 模型公式 ``` ZScore = 0.2086*x1 + 4.3465*x2 + 4.9601*x3 x1: market value/book value of TotalLiabilities x2: total sales/TotalAssets x3: (TotalAssets-TotalAssets[last])/TotalAssets[last] coef = [0.2086, 4.3465, 4.9601] ``` 劃分區間 + `Z-score < 1.5408`: 已經違約 + `Z-score > 1.5408`: 財務健康，短期內不會出現違約情況 ## 3. 如何實用本模塊提供的 Zscore Model 接口 本模塊目前提供了四個Z-score模型接口，分別如下： ### 3.1 `zscore_ACB` ``` --- Interface for calculating zscore using ACB. parameter: ticker[string]: ticker code parameter[opt]: begin[int]: year begin, default 2010 end[int]: year end, default 2014 coef[list of int]: coefficients for Zscore Model, default [0.517, -0.460, 18.640, 0.388, 1.158] --- when success, will return factors and status code, as a dict; when failed for some reason, will return error message and error code, as a dict; --- Model ACB: All Corporate Bonds ZScore = 0.517 - 0.460*TotalLiabilities/TotalAssets + 9.320*NetProfit/0.5*(TotalAssets+TotalAssets[last]) + 0.388*WorkingCapital/TotalAssets + 1.158*RetainedEarnings/TotalAssets = 0.517 - 0.460*x1 + 9.320*2/x2 + 0.388*x3 + 1.158*x4 coef = [0.517, -0.460, 18.640, 0.388, 1.158] ``` ### 3.2 `zscore_ACB_List` ``` --- Interface for calculating zscore using ACB. parameter: ticker[list]: ticker code parameter[opt]: begin[int or list]: year begin, default 2010 end[int or list]: year end, default 2014 coef[list of int]: coefficients for Zscore Model, default [0.517, -0.460, 18.640, 0.388, 1.158] --- ``` ### 3.3 `zscore_ACBEL` ``` --- Interface for calculating zscore using ACBEL. parameter: ticker[string]: ticker code parameter[opt]: begin[int]: year begin, default 2010 end[int]: year end, default 2014 coef[list of int]: coefficients for model, default [0.2086, 4.3465, 4.9601] --- when success, will return factors and status code, as a dict; when failed for some reason, will return error message and error code, as a dict; --- Model ACB: All Corporate Bonds with Equity Listings ZScore = 0.2086*x1 + 4.3465*x2 + 4.9601*x3 x1: market value/book value of TotalLiabilities x2: total sales/TotalAssets x3: (TotalAssets-TotalAssets[last])/TotalAssets[last] coef = [0.2086, 4.3465, 4.9601] --- ``` ### 3.4 `zscore_ACBEL_List` ``` --- Interface for calculating zscore using ACBEL. parameter: ticker[list]: ticker code parameter[opt]: begin[int or list]: year begin, default 2010 end[int or list]: year end, default 2014 coef[list of int]: coefficients for Zscore Model, default [0.2086, 4.3465, 4.9601] --- ``` ## 4. 一個 Zscore Model 實例 相關步驟說明如下： + 4.1 導出該模塊 + 4.2 輸入股票代碼，可選擇性輸入計算時間區間，以年為單位 + 4.3 得到計算結果 + 4.4 作圖分析 ```py # 4.1 & 4.2 GZMT = zscore_ACB('600519') # 4.3 factors = GZMT['factors'] # 4.4 from matplotlib.pylab import plot plot(factors['zscore']) [<matplotlib.lines.Line2D at 0x4eda750>] ```  ## 5. 本模塊未來版本規劃 為適應光大金融人士更加方便地使用 Z-Score Model，本模塊將在以下幾個方面對本模塊的未來版本進行規劃： + 更加簡單：提供上傳模板，用戶只需按照一定格式上傳自己持倉的 excel 表格，即可在零編碼的情況下得到相應持倉債券、股票發行人的 Z-Score 情況； + 更加靈活：將會允許用戶在我們定義的兩個 Z-Score 模型的相關系數； + 更加智能：預計提供給用戶將近50個公司財務相關的因子，由用戶自定義因子和相關的系數來計算 Z-Score 值；